A Reservation and Reuse Combined Q-learning Semi Persistent Scheduling for C-V2X Communication

The 3rd Generation Partnership Project (3GPP) mode 4 provides a direct communication mode to support the C-V2X applications. However, the dynamic traffic load and vehicle mobility lead to uncertainty of channel quality with serious packet collision problem. In order to meet the demand of ultra reliability and low latency of V2X communication. A Reservation-Reuse Combined Q-learning Semi Persistent Scheduling (RRCQSPS) algorithm is proposed for efficient distributed resource allocation in dynamic load environment. Firstly, the theoretical model of the collision probability of Semi Persistent Scheduling (SPS) algorithm is built. Then, the Q-learning model of vehicle agent in dynamic load environment is proposed with the reservation-reuse combined action and Q function. By using -greedy method, the optimal reservation and reuse of wireless resources in dynamic load environment can be solved. The simulation results show that compared with the existing Lookahead-SPS optimization algorithms, the packet reception ratio of RRC-QSPS is improved by 7% and the update packet delay is reduced by 10% in high speed and high load scenarios.